Sunday, August 9, 2026 For AI agents: /llms.txt
AI news for founders, builders and the agents they ship.

Editorial standards, and how they are checked

Every publication claims standards. This page says which of ours a machine verifies on every edition, which still need a person, and what our own records show.

What runs on every edition

After each edition is built, 7 rules are checked against every brief and the result is written to a dated record. The record names the exact version of the style guide it applied, and binds to each brief by the same sha256 content hash the edition JSON publishes, so a verdict cannot follow a story that was edited afterwards.

RuleEffectHow it is checked
No em dashes or en dashes anywhere. Blocks Checked automatically on every edition
Sentence-case headlines, no period. Blocks Checked automatically on every edition
Every fact traces to a named, linked source. Blocks Checked automatically on every edition
Sources are links a reader can click. Blocks Checked automatically on every edition
Original briefs synthesize at least two sources when possible. Flags Checked automatically on every edition
Ledes under 35 words. Flags Checked automatically on every edition
No AI-pattern writing (delve, landscape, leverage, robust, seamless, game-changer, ecosystem, 'not just X, it's Y', 'the reality is'). Flags Checked automatically on every edition

What a machine cannot check

These are the rules that matter most, and they are the ones no automated check can honestly claim. They are recorded as unchecked rather than left off the list, because a scorecard that only lists what is easy to measure reads as a clean bill of health it has not earned.

RuleEffectHow it is checked
Never write a fact you did not see in a source. Blocks A person, or not yet. Requires reading each source and comparing it to each claim.
Quotes attributed with 'said'; unverified claims marked 'X reported' or 'according to X'. Blocks A person, or not yet. Requires knowing which claims were independently verified.
Reporting, not opinion. The last paragraph may say why it matters, without hype. Flags A person, or not yet. Requires reading for stance.
Consequential for founders, builders and developers. No event promos or content marketing. Flags A person, or not yet. Requires editorial judgment about the day's wire.

What a pass does not mean

A clean record means no automated rule was broken. It is not a statement that the reporting is accurate, that the sources support the claims, or that the story was worth running. The check that would tell you that is the one marked as needing a person.

Why the records are not published yet

We keep a record for every edition and we are not publishing them today. Publishing a measurement turns it into a target: the cheapest way to clear a two-source rule is to staple a second link onto a single-source story, which moves the number and improves nothing. We would rather fix the sourcing than decorate the scorecard.

We will publish the records once the gate includes the judgment checks above, so that what you are reading is an audit rather than a spell-check with a certificate. Until then this page is the honest version: here is exactly what we verify, and here is what we do not.

Corrections

If we get something wrong we fix the story and note it on the page. Send corrections to support@nuralyn.com. See also our about page for sourcing and attribution policy.